Digital Microfluidic Platform for the Detection ofRubella Infection and Immunity:

A Proof of ConceptAlphonsus H.C. Ng,1,3 Misan Lee,2,3 Kihwan Choi,3,4 Andrew T. Fischer,5 John M. Robinson,6

and Aaron R. Wheeler1,3,4*

BACKGROUND: Whereas disease surveillance for infec-tious diseases such as rubella is important, it is critical toidentify pregnant women at risk of passing rubella totheir offspring, which can be fatal and can result in con-genital rubella syndrome (CRS). The traditional central-ized model for diagnosing rubella is cost-prohibitive inresource-limited settings, representing a major obstacleto the prevention of CRS. As a step toward decentralizeddiagnostic systems, we developed a proof-of-concept dig-ital microfluidic (DMF) diagnostic platform that pos-sesses the flexibility and performance of automated im-munoassay platforms used in central facilities, but with aform factor the size of a shoebox.

METHODS: DMF immunoassays were developed with in-tegrated sample preparation for the detection of rubellavirus (RV) IgG and IgM. The performance (sensitivityand specificity) of the assays was evaluated with serumand plasma samples from a commercial antirubellamixed-titer performance panel.

RESULTS: The new platform performed the essential pro-cessing steps, including sample aliquoting for 4 parallelassays, sample dilution, and IgG blocking. Testing ofperformance panel samples yielded diagnostic sensitivityand specificity of 100% and 100% for both RV IgG andRV IgM. With 1.8 �L sample per assay, 4 parallel assayswere performed in approximately 30 min with �10%mean CV.

CONCLUSIONS: This proof of concept establishes DMF-powered immunoassays as being potentially useful for thediagnosis of infectious disease.© 2014 American Association for Clinical Chemistry

Vaccine-preventable infectious diseases such as measles,mumps, and rubella continue to be a global threat be-cause many developing countries have low or nonexistentimmunization coverage (1 ). Although rubella is manage-able for most adults and minors (rash, fever, and flu-likesymptoms), rubella infection during pregnancy has sig-nificant risk of causing fetal death and/or a debilitatingsuite of defects known as congenital rubella syndrome(CRS)7 (2 ); �100000 infants are born with CRS eachyear (3 ).

As is the case for many infectious diseases, screeningfor rubella susceptibility, immunity, and infection ischallenging, requiring the selective detection of rubellavirus (RV)-specific IgG and IgM. Because the test com-ponents of RV IgM immunoassays can cross-react withother biogenic proteins [e.g., RV IgG or rheumatoid fac-tor (RF) IgM], there is a high occurrence of false-negativeor false-positive results, depending on the quality of theassays and the experience of the operators (4–6 ). Thus,although there are portable tests available (7 ), rubella isalmost always diagnosed in central clinical laboratories.In resource-limited settings, transportation from ruralcommunities to remote laboratories is often cost prohib-itive (8 ), meaning that rubella infections (and attendantcases of CRS) routinely sweep through rural populationswithout testing or treatment (9 ).

Given the challenges described above, there is greatenthusiasm for moving to decentralized systems fordisease diagnosis and surveillance. This model requiresportable assays, and the microfluidic (or “lab on achip”) community has attempted to address this chal-lenge primarily through systems that rely on fluid flowthrough enclosed microchannels (10–12) or lateral flowthrough a paper-based absorptive matrix (13–15).Microchannel-based systems have excellent analyticalsensitivity (16 ) and proven diagnostic performance in

1 Institute of Biomaterials and Biomedical Engineering, 2 Innis College, and 4 Departmentof Chemistry, University of Toronto, Toronto, ON, Canada; 3 Donnelly Centre for Cellularand Biomolecular Research, Toronto, ON, Canada; 5 Abbott Diagnostics, Irving, TX; 6 Ab-bott Diagnostics, Abbott Park, IL.

* Address correspondence to this author at: University of Toronto, 80 St. George St. To-ronto, ON, Canada M5S 3H6. Fax 416-946-3865; e-mail aaron.wheeler@utoronto.ca.

Received August 27, 2014; accepted November 13, 2014.

Previously published online at DOI: 10.1373/clinchem.2014.232181© 2014 American Association for Clinical Chemistry7 Nonstandard abbreviations: CRS, congenital rubella syndrome; RV, rubella virus; RF,

rheumatoid factor; DMF, digital microfluidics; Cal, calibrator; HRP, horseradish peroxi-dase; PMT, photomultiplier tube; LOD, limit of detection; LOQ, limit of quantification;s/co, signal-to-cutoff ratio; EIA, enzyme immunoassay; RLU, relative light unit.

Clinical Chemistry 61:2420–429 (2015)

Automation and Analytical Techniques

420

resource-limited settings (17–19), but often require ex-tensive ancillary equipment to operate (e.g., pumps, flowmeters, and valves) and intricate device fabrication forintegrated sample preparation (20 ). In contrast, paper-based systems can be simple and inexpensive to manufac-ture and use (21 ), but adequate analytical performance insuch devices has been demonstrated for only a few ana-lytes (15 ), and multistep liquid manipulation with uni-form flow profiles requires careful material manipulationand device design (22–24).

An alternative to (channel-based or lateral) fluidflow for miniaturized analysis systems is a technologyknown as digital microfluidics (DMF) (25–27), in whichsample and reagents are manipulated as discrete dropletson a hydrophobic surface. DMF systems actuate dropletsthrough the application of electrical potentials on a ge-neric array of insulated electrodes—a format that enablessoftware-reconfigurable, concurrent droplet operationsincluding merging, mixing, splitting, and metering fromreservoirs (28 ). DMF devices can be cost-effectively fab-ricated on paper by inkjet printing (29 ) and can beoperated with simple and compact instrumentation(30, 31 ) with no need for pumps, interconnects, valves,or fittings. Although these advantages have motivated thedevelopment of DMF-powered immunoassays for small-molecule and protein biomarkers (32–35), we are notaware of DMF immunoassays being used for diagnosis ofinfectious diseases.

Here, we report the development of DMF-poweredRV IgG and RV IgM immunoassays, with a focus onevaluating the diagnostic performance of DMF immuno-assays relative to central laboratory methods. These assaysuse RV-immobilized magnetic particles to capture theanalyte from the sample and enzyme-linked detectionantibodies to transduce analyte-binding events to chemi-luminescent signal. All assay steps, from sample to anal-ysis, were performed by DMF with a shoebox-sized pro-totype instrument. We propose that this represents auseful step toward the development of decentralized di-agnostic tools for the diagnosis of rubella and other in-fectious diseases.

Materials and Methods

REAGENTS AND MATERIALS

Unless otherwise specified, we purchased reagents fromSigma-Aldrich. Deionized water had a resistivity of 18M� � cm at 25 °C. Pluronic L64 (BASF Corp.) was gen-erously donated by Brenntag Canada.

We adapted most immunoassay reagents from theArchitect rubella IgM and IgG assay kits from AbbottLaboratories, including RV IgG calibrators, RV IgM cut-off Calibrator 1 (Cal 1), RV IgM pretreatment reagentcontaining goat antihuman IgG, and virus-coated para-magnetic microparticles. Reagents from other vendors in-

cluded Superblock Tris-buffered saline and SuperSignalELISA Femto chemiluminescent substrate, comprisingstable peroxide (H2O2) and luminol-enhancer solution,from Thermo Fischer Scientific; antirubella mixed-titerperformance panel (PTR201-00-0.5) from SeraCare LifeSciences; and defibrinated sheep blood from Quad Five.

We prepared custom digital microfluidic compati-ble wash buffer and conjugate diluent as described previ-ously (32, 33 ). Conjugate working solutions for RV IgGor RV IgM assays were formed by diluting horseradishperoxidase (HRP)-conjugated goat polyclonal anti-human IgG (21 ng/mL) or HRP-conjugated goat anti-human IgM (23 ng/mL), respectively, in conjugate di-luent. Pretreatment working solution was formed from a3� dilution of IgM pretreatment reagent in conjugatediluent. The microparticle working suspension wasformed in an Eppendorf tube by removing the originaldiluent (with particles immobilized with a neodymiummagnet), washing the particles twice in Superblock, andresuspending the washed particles in Superblock at 1.5 �108 particles/mL (approximately 10� the stock concen-tration). Simulated whole blood samples were formedfrom a 1:1 mixture of RV IgG calibrator and sheep blood.Before running immunoassays, we diluted calibrators,simulated whole-blood samples, or patient samples (fromperformance panel) 10� in Dulbecco PBS containing4% BSA. All reagents used on digital microfluidic devicesor on-chip were supplemented with Pluronic L64 at0.05%.

DEVICE FABRICATION AND OPERATION

We used a custom automation system with approximatedimensions 7 � 9 � 12 inches to manage droplet oper-ation, magnet and photomultiplier tube (PMT) position,and data collection (Fig. 1A) (30, 32 ). Digital microflu-idic devices, comprising a top plate and bottom plate,were fabricated in the University of Toronto Nanofabri-cation Centre cleanroom facility and assembled as de-scribed previously (32 ) (Fig. 1B). The bottom plate de-vice design featured an array of 80 actuation electrodes(2.54 � 2.54 mm) connected to reservoir electrodes forsample and reagent storage and waste removal. Unitdroplet volumes on the actuation electrodes were approx-imately 900 nL, as determined by the area of each actu-ation electrode and the gap spacing (approximately 140�m) between the top plate and bottom plate.

Droplets were actuated by applying a prepro-grammed sequence of driving voltages (80–100 VRMS 10kHz sine wave) between the top plate (ground) and elec-trodes in the bottom plate through a Pogo pin interface(90 pins). For on-chip particle separation, a motorizedmagnet system called a magnetic lens (32 ) is positionedapproximately 150 �m underneath the device. In thisactivated state, �600 �N of magnetic force is sufficientto focus particles into a pellet, immobilizing them on the

Microfluidic Rubella Assay

Clinical Chemistry 61:2 (2015) 421

surface (Fig. 1C; also see Supplemental Video 1, whichaccompanies the online version of this article at http://www.clinchem.org/content/vol61/issue2). With digitalmicrofluidic actuation, the supernatant can be separatedfrom the particles, provided that the immobilizationforce exceeds the minimal threshold (approximately 500�N) required to overcome the surface tension of thedroplet (32, 33 ). The particle pellet is resuspended bydeactivating the magnetic lens (i.e., magnet is positioned3.8 cm below the device) and moving and mixing a drop-let over the particles (Fig. 1D).

To measure on-chip chemiluminescence, the reac-tion droplet is moved to the center of the electrode arrayand a motorized PMT is positioned several hundredmicrometers above the device to collect light from thedroplet. When the PMT is inactive (i.e., the PMT ispositioned 5 cm above the device), an integrated light-emitting diode and webcam assembly is used to monitoron-chip droplet movement. We implemented 3 designmeasures to prevent users from damaging the PMT, in-

cluding (a) 2 sensors to ascertain the position of thePMT, (b) a shutter mechanism to prevent light fromsaturating the sensor when it is not in use, and (c) deepgrooves in the enclosure to ensure that the instrument islight-tight during measurements.

RV IgG IMMUNOASSAY

RV IgG antibodies were detected on-chip with an indi-rect ELISA comprising 7 steps. (a) A droplet of virus-coated paramagnetic particle suspension (1.8 �L) wasdispensed from a reservoir, and the particles were sepa-rated from the diluent. (b) A droplet of sample (1.8 �L,containing RV IgG) was dispensed, delivered to the par-ticles, mixed for 3 min, and separated from particles. (c)The particles were washed in 4 successive droplets ofwash buffer (i.e., 4 � 1.8 �L), each time mixing and thenseparating from the supernatant. (d ) A droplet of HRPconjugate solution (1.8 �L, containing HRP-conjugatedantihuman IgG) was dispensed, delivered to the particles,and mixed for 2 min. (e) Step c was repeated. (f ) One

23cm

30cm

18cm

PMT

Magnetic lens

PMT

Magneticlens

g

Electrical interfaceto device

A

Front view (photograph) Cross-section view (rendering)

Electrical inputs

Motor

Motors

Top plate

Bottom plate

Droplet

~

AirGlassElectrodeHydrophobicInsulator

140 µm

Magnetic lens

Magnetic lens

B2 in

3 in

Top plate

Bottomplate

Electrodearray

Interface toinstrument

CParticle

suspensionIgM IgG

2.54 mm

1 2

3 4

Particle pelletSeparation zone

Magnetic lens active

Waste

D1 2 3

4 5 6

7 8 9

Fig. 1. Automated digital microfluidic platform.(A), Left: Photo of automation system (front view) showing the position of electrical inputs from the DropBot (30 ) and external motors thatcontrol the PMT and enclosure. Right: Cross-section computer-aided design rendering showing the interior components of the system,including the PMT, the magnetic lens, and the motor system that controls the position of the magnetic lens. (B), Three-dimensional schematicand side view of digital microfluidic device positioned above the magnetic lens. (C), Video sequence of digital microfluidic magnetic separa-tion of paramagnetic particles in parallel RV IgM and RV IgG immunoassays (see online Supplemental Video 1). (D), Video sequence ofresuspending and mixing of particles in wash buffer solution. Red arrows indicate the direction of droplet movement.

422 Clinical Chemistry 61:2 (2015)

droplet each of H2O2 (900 nL) and luminol-enhancersolution (900 nL) were dispensed and merged with theparticles; the solutions and particles were mixed for 2min. (g) The mean chemiluminescent signal over 10 s wasobtained with the integrated PMT.

By aliquoting 4 sets of samples and reagents (seeonline Supplemental Fig. 1), 4 parallel tests were runsimultaneously (i.e., steps a–g were performed in parallelfor 4 samples). Applying this protocol, a calibration curvefor RV IgG was established in 4 replicates (intraassay),with calibrator solutions containing 0, 5, 10, 15, 75, and250 IU/mL RV IgG. The limits of detection (LOD) andquantification (LOQ) for this assay were the concentra-tions corresponding to the position on the calibrationcurve of the mean signal generated from blank measure-ments plus 3 (LOD) or 10 (LOQ) times the SD of theblank measurements. A set of calibration data was usedfor repeated experiments until a new batch of devices ormagnetic particles were introduced, or when the controlsample measurements were observed to be out of range.In practice, this allowed the use of each set of calibrationdata for �200 experiments over 50 days.

INTEGRATED ON-CHIP DILUTION AND RV IgG IMMUNOASSAY

As shown in online Supplemental Fig. 2, samples werediluted on-chip in 3 iterating steps. (1) A droplet of sam-ple (1.8 �L) and a droplet of diluent (1.8 �L, DulbeccoPBS 4% BSA) were dispensed, merged, and mixed. (2)The pooled sample droplet (2� diluted) was split into 2subdroplets. (3) One sample subdroplet was stored forlater analysis, and the second sample subdroplet was usedto repeat steps 1–3 until the desired dilutions wereachieved. For example, 4 repetitions generate 4 dropletswith 2�, 4�, 8�, and 16� dilution of original sample.After dilution, RV IgG immunoassays were performed,with the 7-step (a–g) procedure above, by delivering thediluted samples to virus-coated particles. When the par-ticles were washed, the regions on the device used fordilution and sample storage were also washed with addi-tional wash buffer to prevent cross-contamination.

A fully integrated protocol, including dilution in 4iterations of steps 1–3 followed by RV IgG immunoas-says in steps a–g, was evaluated with a RV IgG calibratorand a simulated whole-blood sample (1:1 mixture of RVIgG calibrator and sheep blood), both containing 250IU/mL of RV IgG. Here, on-chip dilution was per-formed to generate droplets containing 15.6, 31.3, 62.5,and 125 IU/mL RV IgG. For each sample (calibrator andwhole blood), the protocol was repeated on 3 differentdevices (i.e., 3 interassay replicates).

OFF-CHIP RF IgM SCREENING

We used an RF IgM ELISA (Orgentec Diagnostika) ac-cording to manufacturer’s instructions to screen for RFIgM in the RV IgG calibrators and RV IgM Cal 1. In

these experiments, at least 2 replicates were evaluated foreach sample by absorbance, measured at 450 nm (refer-ence 650 nm) with a Sunrise microplate reader (Tecan) in“Accuracy” read mode. We estimated the RF IgM con-centrations of the samples from a calibration curve, gen-erated by fitting the absorbance of the ELISA calibratorsolutions to a 4-parameter logistic equation.

RV IgM IMMUNOASSAY

The DMF RV IgM immunoassay was identical to the7-step protocol (a–g) described above, except that theconjugate solution contained HRP-conjugated anti-human IgM (instead of IgG) and the sample was pre-treated with antihuman IgG (before step a). In the pre-treatment procedure, a droplet of sample (1.8 �L) and adroplet of pretreatment reagent (0.9 �L) were mergedand mixed for 7 min. The assay was evaluated with andwithout pretreatment with 4 test samples: (i) 0 IU/mLRV IgG calibrator (control), (ii) 250 IU/mL RV IgGcalibrator (IgG), (iii) 1:1 mixture of Cal 1 and 0 IU/mLRV IgG calibrator (IgM), and (iv) 1:1 mixture of Cal 1and 500 IU/mL RV IgG calibrator (IgM � IgG). Foreach sample, 4 intraassay replicates were performed foreach condition tested. The results (for IgG, IgM, andIgG � IgM) were reported as fold change, which wasobtained by normalizing the signal intensity of each assayto the average signal intensity of the respective controlexperiments.

Applying the integrated RV IgM protocol (includ-ing both pretreatment and immunoassay), we establishedthe cutoff calibration data in 3 replicates with unmodi-fied Cal 1. In addition, the background signal of the RVIgM assays was evaluated in 4 replicates (intraassay) witha 0 IU/mL RV IgG calibrator.

ANTIRUBELLA MIXED-TITER PERFORMANCE PANEL

We estimated the diagnostic accuracy of DMF RV IgGand RV IgM immunoassays with 25 plasma/serum sam-ples (PTR201-01-25, SeraCare Life Sciences). Serum orplasma samples were thawed and aliquoted into single-use vials for DMF immunoassays. Working solutions in-cluding microparticles, conjugates, and chemilumines-cent substrates were made fresh each day from stocksolutions.

DMF immunoassay analyses for the panel sampleswere completed by 3 different operators over a 1-monthperiod. Each assay was carried out in 3–4 replicates inparallel, and operators were blinded to the reference re-sult until all of the data were collected. The DMF assayresults were reported as signal-to-cutoff ratio (s/co) val-ues, calculated by normalizing the mean signal intensityof each assay to the mean signal intensity of the respectivecutoff control. The cutoff controls for RV IgG and RVIgM assays were the 10 IU/mL RV IgG calibrator and theRV IgM cutoff Cal 1 (both included in the respective

Microfluidic Rubella Assay

Clinical Chemistry 61:2 (2015) 423

kits), respectively. A DMF s/co �1 was interpreted aspositive, a DMF s/co in the gray zone (0.69–0.99 forIgG and 0.75–0.99 for IgM) was interpreted as equivo-cal/indeterminate, and a DMF s/co below the gray zone(�0.69 for IgG and �0.75 for IgM) was interpreted asnegative.

EVALUATION OF DIAGNOSTIC ACCURACY

We calculated the diagnostic sensitivity and specificityfor the DMF immunoassays with the Abbott enzymeimmunoassay (EIA) as a reference testing method (36 ).For cases in which the calculated sensitivity or specificitywas equal to 1, a P value of 0.999 was used to calculate95% CIs, as recommended by WHO (37 ) and others(17, 18 ).

Results and Discussion

DIGITAL MICROFLUIDIC INSTRUMENTATION

Motivated by the need to identify patients at risk forCRS, we developed DMF immunoassays for RV-specificimmunoglobulins. The challenging nature of this appli-cation, which includes the need to distinguish betweenRV IgG and RV IgM, necessitates a sophisticated systemcapable of handling a large number of reagents and per-forming a complex series of processing steps. To achievesuitable levels of processing sophistication in a miniaturizedinstrument, we adapted an in-house designed computer-controlled instrument (32 ) powered by DropBot (30 ),an open-source DMF automation system.

The shoebox-sized instrument comprises a motor-ized PMT for chemiluminescent detection, an electronicinterface to the DropBot for droplet movement, and amagnet assembly (called a magnetic lens) for particle sep-aration (Fig. 1A). Movement of protein-containingdroplets through an air-filled device (Fig. 1B), which al-lows for more efficient particle-concentration and ana-lyte washing than the more common methods relying onoil-filled devices (33 ), is facilitated by the inclusion ofPluronic additives in samples and reagents, which re-duces protein adsorption and limits cross-contamination(38 ). During particle separation, particles in the dropletare immobilized on the surface in a region above themagnet, referred to as the separation zone (Fig. 1C; on-line Supplemental Video 1). Subsequently, the droplet isactuated away to waste, leaving behind a particle pellet.For particle resuspension, the magnetic lens is deacti-vated, and a new droplet is actuated over the particlepellet and mixed in a circular motion; this is sufficient tobreak up the particle pellet and completely reconstitutethe particles in suspension (Fig. 1D).

The DMF instrumentation and devices describedhere are prototype designs used to demonstrate the diag-nostic performance of DMF immunoassays. In ongoingwork, we are developing next-generation systems that

may be suitable for portable testing in the field [i.e., lowcost, user friendly, battery-powered, and compact (39 )].These new designs will likely benefit from the develop-ment of low-cost device fabrication methods (29 ), simpleinstrumentation (30, 31 ), and integrated electrochemi-cal sensors (40, 41 ).

RV IgG IMMUNOASSAY AND SAMPLE DILUTION

Whether acquired naturally or induced by vaccination,the persistence of RV IgG antibody confers protectionagainst rubella infection. The WHO interpretation ofRV IgG serological results is dependent on the antibodyconcentration: a sample is positive if antibody concentra-tion is �10 IU/mL, equivocal if it is between 5 and 10IU/mL, and negative if it is �5 IU/mL. Individuals withantibody concentrations �10 IU/mL are susceptible toclinical illness on exposure to rubella virus. As such, test-ing women for RV IgG before conception or at their firstantenatal visit can help minimize the risk of CRS.

Using the particle separation and resuspension tech-niques described above, we developed a 7-step DMF in-direct ELISA for RV IgG, in which samples or reagentsare aliquoted to the particles for 4 parallel assays (seeonline Supplemental Fig. 1). As shown in Fig. 2A, RVIgG antibodies present in the sample bind to the RV-coated paramagnetic particles. After incubation andwashing, HRP-conjugated antihuman IgG is added toform immunocomplexes. After another incubation andwash cycle, hydrogen peroxide and luminol-enhancer so-lutions are added to the particles, and the resultingchemiluminescence is measured as relative light units(RLUs). This assay procedure, including determinationof diluent compositions, reagent concentrations, incuba-tion times, droplet operations, and chemiluminescentsignal generation, was developed from design of experi-ment optimization (32 ). The use of chemiluminescenceis particularly advantageous because few optical compo-nents are required (e.g., no excitation source, filters, orlenses).

The assay described here was designed for detectionof RV in patient serum and plasma samples. Thus, tooptimize and calibrate the method, we used commercialcalibrators designed to simulate human serum (buffercontaining 4% BSA). As shown in the calibration curve(Fig. 2B), a direct relationship exists between the RLUsdetected by the integrated PMT and the amount of RVIgG in the calibrators. The calibration curve had a dy-namic range of approximately 2 orders of magnitude, theintraassay CVs ranged from 0.3% to 6.5%, and the LODand LOQ of the assay were 0.03 and 0.1 IU/mL, respec-tively, below the clinical cutoff concentrations describedabove. Each set of 4 assays requires approximately 25 minto complete.

In some instances, samples with high analyte con-centration must be diluted before analysis to facilitate

424 Clinical Chemistry 61:2 (2015)

accurate quantification; this technique is relatively trivialfor central laboratories, but is not commonly used inportable tests. Hence, we evaluated whether the DMFplatform could reliably perform a range of dilutions onsamples containing 250 IU/mL RV IgG, which was be-yond the linear range of the calibration curve (Fig. 2B). Itis straightforward to dilute such samples on-chip—forexample, to form 4 droplets of 125, 62.5, 15.6, and 31.3IU/mL (see online Supplemental Fig. 2). Subsequently,RV IgG immunoassays can be performed on the dilutedsamples in parallel.

The assays described here were designed for applica-tion to patient serum and plasma samples. But for porta-ble analysis in the field, it would be useful if the assaycould also be used with whole blood, negating the re-quirement of phlebotomy (with finger stick) and frac-tionation. To evaluate this possibility, simulated wholeblood samples (calibrator mixed with sheep’s blood) werediluted on-chip and tested with the methods describedabove. As shown in Fig. 2C, the interassay CVs for theseassays were only marginally higher in the blood samples(10% to 15%) than in the control (standard) samples(6% to 9%). Because of dissimilarities in matrix compo-

sition, the signal intensities of the calibrator samples dif-fered significantly from the diluted samples containinghigh blood content (i.e., P � 0.049 for the 1/8 dilutionand P � 0.050 for the 1/4 dilution). But when thesesamples were diluted further (i.e., 1/16 and 1/32 dilu-tions of blood), the results for blood and calibrator sam-ples could not be distinguished (P � 0.05). These obser-vations suggest that sample dilution may be useful foralleviating interferences arising from the blood matrix forapplication in the field, provided that the dilution doesnot reduce the analyte concentration below the LOQ ofthe assay.

RV IgM IMMUNOASSAY AND IgG BLOCKING

Whereas testing for RV IgG (described above) is usefulfor determining susceptibility and immune status, thecritical test for diagnosing rubella-infected patients (in-cluding the identification of pregnancies at risk for CRS)is the detection of RV IgM. Unlike RV IgG, there is noagreed cutoff concentration for RV IgM to distinguishbetween negative or positive test results. As such, assayvendors develop their own cutoff calibrators on the basisof serological testing of samples obtained from healthy

Sample (1.8 µL)

Incubate (3 min)

Conjugate (1.8 µL)

Incubate (2 min)Wash (4 )

H2O2 (0.9 µL)Luminol (0.9 µL)

Incubate (2 min)Enzymatic

chemiluminescentreaction

Rubella virus on magnetic particle Rubella virus IgG

Peroxidase-conjugatedantihuman IgG

A

B

4 8 16 32 64 128 2560

2.0 ××106

4.0 ××106

6.0 ××106

8.0 ××106

1.0 ××107

Diagnostic cutoff(10 IU/mL)

Rubella virus IgG (IU/mL)

Sign

al (R

LU)

C

15.6 31.3 62.5 1250

4.0 ××106

8.0 ××106

1.2 ××

××Wash (4 )××

107

Rubella virus IgG (IU/mL)Si

gnal

(RLU

)

BloodStandard

P = 0.09

P = 0.08

P < 0.05*

P < 0.05*

1/4

Blo

od

1/8

Blo

od

1/16

Blo

od

1/32

Blo

od

Fig. 2. RV IgG immunoassay.(A), Scheme and procedure of RV IgG immunoassay on paramagnetic particles. (B), Calibration data (red diamonds) and curve (dotted line) forRV IgG immunoassay implemented in the digital microfluidic system. (C), Comparison of RV IgG immunoassay results for samples in calibratorsolution (gray) and blood (white) for dilutions formed on-chip. Each value represents the mean (SD) from 3 replicates. Statistical significance(*P < 0.05) was evaluated by Student 2-tailed t-test.

Microfluidic Rubella Assay

Clinical Chemistry 61:2 (2015) 425

and infected donors. In the assays reported here, the RVIgM Cal 1 reagent was adapted from the Abbott Archi-tect rubella IgM assay kit to define the cutoff betweenpositive and negative/equivocal results with digitalmicrofluidics.

Similar to the RV IgG immunoassays describedabove, the on-chip RV IgM assays relied on RV-immobilized paramagnetic particles to implement indi-rect ELISA. In an ideal assay, RV IgM antibodies fromthe sample will bind to the particles, and the detectionantibodies (HRP-conjugated antihuman IgM) will bindto the RV IgM to form immunocomplexes (Fig. 3A,panel 1). However, there are at least 2 potential sources oferror in this assay scheme caused by the presence of RVIgG and/or RF IgM in human serum (42 ). If the samplecontains RV IgG, these antibodies will compete for bind-ing sites on the particles, leading to false-negative results(Fig. 3A, panel 2). If the sample contains both RV IgGand RF IgM, immunocomplexes will form on the parti-cles, leading to false-positive results (Fig. 3A, panel 3).

Complicating matters, vendor-specific calibrators, whichare formed from pooled human sera, often contain vary-ing amounts of RF IgM or other cross-reactive species,depending on the source. To ascertain the presence of RFIgM in the calibrators used here, they were tested with anoff-chip ELISA kit (see online Supplemental Table 1)—the RV IgG calibrators have low concentrations of RFIgM (1.4–1.85 IU/mL), and the RV IgM cutoff calibra-tor has high concentrations of RF IgM (27.0 IU/mL).

Fortunately, errors related to the presence of endog-enous RV IgG and RF IgM (Fig. 3A) can be alleviated bypretreating the sample with exposure to exogenous anti-human IgG before analysis, which serves to block un-wanted binding of RV IgG (and RV IgG/RF IgM com-plexes) onto the virus-laden particles (4–6 ). Emulatingthe Architect rubella IgM assay procedures, we developedan on-chip pretreatment method in which a droplet ofsample (1.8 �L) is mixed with a droplet of pretreatmentreagent (900 nL, containing goat antihuman IgG) for 7min before analysis by on-chip ELISA. To evaluate this

1) RV IgM assay 2) False negative 3) False positive

A

Magnetic particle

Rubella virus (RV)

RV IgM

RV IgG

Detection anti-IgM

Rheumatoid factor IgM

C

00

4.0×105

8.0×105

1.2×106

Sign

al (R

LU)

IgM cutoff

B

IgG IgM IgM + IgG0

5

10

15

20

25No pretreatment

Pretreatment

Fold

chan

ge

P = 0.0003*

P = 0.6

P = 0.5

Fig. 3. RV IgM immunoassay.(A), Scheme of RV IgM immunoassay (panel 1) and potential interferences or nonspecific interactions arising from the presence of RV IgG(panel 2) and RF IgM (panel 3). (B), Bar graph illustrating the effect of on-chip sample pretreatment (gray bars) relative to no pretreatment(white bars) in RV IgM immunoassays. (C), Diagnostic cutoff calibrator signal for RV IgM immunoassays implemented in digital microfluidicswith sample pretreatment. Each value represents the mean (SD) from 3– 4 replicates. Statistical significance (*P < 0.05) was evaluated byStudent 2-tailed t-test.

426 Clinical Chemistry 61:2 (2015)

method, we used test samples (formed from calibrators)containing RV IgG, RV IgM, or both RV IgG and RVIgM. As expected, the IgG-alone and IgM-alone sampleresults were unaffected by the pretreatment (Fig. 3B)because these samples did not contain high concentra-tions of RF IgM and RV IgG (see online SupplementalTable 1). In contrast, the pretreatment significantly (P �0.0003) suppressed the nonspecific signal arising in thecombined IgG plus IgM sample, which, as shown in on-line Supplemental Table 1, contains substantial concen-trations of RV IgG (250 IU/mL) and RF IgM (approxi-mately 14 IU/mL). Thus, accurate diagnosis of rubellainfection requires IgG blocking, which is straightforward

to implement and to establish cutoff values (Fig. 3C) indigital microfluidics.

EVALUATION OF DIAGNOSTIC ACCURACY

The diagnostic accuracies of the digital microfluidic RVIgG and IgM immunoassays were estimated with 25plasma/serum samples from a commercial rubella mixed-titer performance panel (Fig. 4). These are undiluted ali-quots of plasma or serum collected from individual do-nors between 1994 and 1995 by the vendor, used bydiagnostic laboratories and manufacturers to evaluatetheir rubella tests. For each sample, the vendor pro-vided reference results (s/co values and test interpreta-

DM

F ru

bella

IgG

(s/c

o)

0.1

1

10

Abbott EIA rubella IgGPositive Negative

0.0 0.5 1.00.0

0.5

1.0AUC = 1.00

1 − Specificity

1 − Specificity

Sens

itivi

ty

AD

MF

rube

lla Ig

M (s

/co)

0.1

1

10

Abbott EIA rubella IgMPositive Equivocal Negative

0.0 0.5 1.00.0

0.5

1.0AUC = 1.00

Sens

itivi

ty

B

C

Sample DMF s/co(result)

Abbott EIAs/co (result)

PTR201-01 1.3 (pos) 1.1 (pos) 0.2 (neg) 0.2 (neg)PTR201-02 2.0 (pos) 2.4 (pos) 0.7 (neg) 0.5 (neg)PTR201-03 4.6 (pos) 3.1 (pos) 2.8 (pos) >2.3 (pos)PTR201-04 2.2 (pos) 2.8 (pos) 0.1 (neg) 0.2 (neg)PTR201-05 0.2 (neg) 0.1 (neg) 0.2 (neg) 0.1 (neg)PTR201-06 1.2 (pos) 1.2 (pos) 0.2 (neg) 0.1 (neg)PTR201-07 2.7 (pos) 2.5 (pos) 2.5 (pos) 1.2 (pos)PTR201-08 2.0 (pos) 1.7 (pos) 0.1 (neg) 0.2 (neg)PTR201-09 2.4 (pos) 2.7 (pos) 0.2 (neg) 0.1 (neg)PTR201-10 1.7 (pos) 3.0 (pos) 0.1 (neg) 0.1 (neg)PTR201-11 1.7 (pos) 2.7 (pos) 3.4 (pos) >2.3 (pos)PTR201-12 3.1 (pos) 2.7 (pos) 0.5 (neg) 0.2 (neg)PTR201-13 2.4 (pos) 3.2 (pos) 0.2 (neg) 0.7 (neg)PTR201-14 1.1 (pos) 2.4 (pos) 0.5 (neg) 0.1 (neg)PTR201-15 2.7 (pos) 3.3 (pos) 0.4 (neg) 0.1 (neg)PTR201-16 0.2 (neg) 0.1 (neg) 0.4 (neg) 0.2 (neg)PTR201-17 2.0 (pos) 2.0 (pos) 0.5 (neg) 0.2 (neg)PTR201-18 1.9 (pos) 1.1 (pos) 1.3 (pos) 0.9 (equ)PTR201-19 2.2 (pos) 1.7 (pos) 1.7 (pos) 2.1 (pos)PTR201-20 3.2 (pos) >3.4 (pos) 0.2 (neg) 0.1 (neg)PTR201-21 2.7 (pos) 2.2 (pos) 0.3 (neg) 0.2 (neg)PTR201-22 1.1 (pos) 1.4 (pos) 0.5 (neg) 0.1 (neg)PTR201-23 1.6 (pos) 2.2 (pos) 0.3 (neg) 0.1 (neg)PTR201-24 1.5 (pos) 1.3 (pos) 0.8 (equ) 1 (pos)PTR201-25 1.8 (pos) 2.5 (pos) 0.2 (neg) 0.3 (neg)

Rubella virus IgG Rubella virus IgMDMF s/co

(result)Abbott EIAs/co (result)

Result differs from Abbott EIA (determined by supplier)

Fig. 4. Performance of RV IgG and RV IgM immunoassay on commercial plasma/serum panel.Plots of s/co values for DMF RV IgG (A) and RV IgM (B) immunoassays from RV mixed-titer panel (Seracare PTR201). Each DMF s/co valuerepresents the mean from 3– 4 replicates and is categorized as positive (pos), equivocal (equ), or negative (neg) on the basis of results from aleading commercial RV antibody immunoassay (Abbott EIA) determined by the supplier. DMF s/co values in the shaded region were inter-preted as equivocal. The inset plots show the ROC curves for DMF assays (red lines) and random guesses (dashed lines). The area under the ROCcurve (AUC) is reported in each plot. (C), Table comparing the results of the DMF and Abbott EIA assays. The s/co and interpretation of the AbbottEIA assays were supplied by Seracare. Shaded box indicates discrepancy in DMF assay with Abbott EIA.

Microfluidic Rubella Assay

Clinical Chemistry 61:2 (2015) 427

tion) from tests performed by the vendor and indepen-dent reference laboratories (between 1995 and 1996)with several leading ELISA platforms at that time. Inthis study, the Abbott EIA test was chosen as the ref-erence standard because test results were provided forboth RV IgM and IgG.

In estimating diagnostic performance, all 25 sampleswere divided and tested for both RV IgG and RV IgM in3–4 replicates to ascertain assay reproducibility. Typi-cally, 4 parallel immunoassays, requiring approximately25 (for RV IgG assays) or 35 (for RV IgM assays) min,were carried out in single (only RV IgG or RV IgM inquadruplicates) or duplex (both RV IgG and RV IgM induplicates) format. The mean intraassay CV (range) was4.6% (0.4%–17.5%) for RV IgG assays and 7.5%(1.5%–13.4%) for RV IgM assays. In the RV IgG assays(Fig. 4A), the DMF tests identified 23 positives, 0 equi-vocal, and 2 negatives, which represents perfect agree-ment with the Abbott EIA results. The diagnostic sen-sitivity and specificity (with 95% CIs) were 100%(98.6%–100%) and 100% (95.5%–100%), respec-tively (see online Supplemental Table 2). In the RVIgM assays (Fig. 4B), the DMF tests identified 5 pos-itives, 1 equivocal, and 19 negatives, which representsclose agreement with the Abbott EIA results (2 mis-matches: 1 DMF equivocal/Abbott positive and 1DMF positive/Abbott equivocal). In the calculation ofdiagnostic performance, the “equivocal” results (a clas-sification not accounted for) were interpreted as “pos-itive,” yielding a sensitivity and specificity (with 95%CIs) of 100% (97.4%–100%) and 100% (98.4%–100%),respectively (see online Supplemental Table 3).

The measured s/co for all 25 samples tested areshown in Fig. 4C. For the 2 mismatches in the RV IgMassay, the DMF values (s/co � 0.9 and 1) were at ornear the cutoff value (s/co � 1). Thus, we hypothesizethat the discrepancy between the DMF and the stan-dard results may be caused by a combination of assayvariability (CV �15%) and degradation of the 20-year-old samples. Regardless, these types of positive-equivocal discrepancies represent a relatively minorproblem, as in standard practice, all patients who testeither positive or equivocal are retested to ascertainfinal diagnosis (43 ).

The performance of the new DMF test (Fig. 4) wassimilar to that of central laboratory–based RV testing

(44 ). This is not surprising, since DMF, like automatedimmunoassay platforms, is automated and software pro-grammable, allowing for integration of operations such asaliquoting, dilution, or IgG blocking as needed in animmunoassay work flow. More importantly, when com-pared with the only “rapid” rubella IgM test that we areaware of (Alere ImmunoComb® Rubella IgM), forwhich the manufacturer reports diagnostic sensitivityand specificity [sensitivity/specificity 87%/99% (7 )], theDMF method has the added value of built-in, automatedreplicates and digital readout (i.e., no user interpretationof results). Of course, to further validate the diagnosticperformance of the new method, additional testing mustbe performed in countries such as Vietnam, where rubellahas not been introduced into the national immunizationprogram (45 ). Nevertheless, these initial results suggestthat a portable DMF-based system, perhaps combinedwith inexpensive, paper devices (29 ), may represent auseful new tool for identification of patients at risk forCRS in low-resource settings. More generally, we pro-pose that similar methods might be useful for distrib-uted diagnostics for a wide range of infectious diseases.

Author Contributions: All authors confirmed they have contributed tothe intellectual content of this paper and have met the following 3 require-ments: (a) significant contributions to the conception and design, acquisi-tion of data, or analysis and interpretation of data; (b) drafting or revisingthe article for intellectual content; and (c) final approval of the publishedarticle.

Authors’ Disclosures or Potential Conflicts of Interest: Upon man-uscript submission, all authors completed the author disclosure form. Dis-closures and/or potential conflicts of interest:

Employment or Leadership: None declared.Consultant or Advisory Role: None declared.Stock Ownership: A. Fischer, Abbott.Honoraria: None declared.Research Funding: This study was supported by the Natural Sciencesand Engineering Research Council of Canada (NSERC) and AbbottDiagnostics. A. Ng, graduate fellowship from NSERC; A.R. Wheeler,Canada Research Chair Program.Expert Testimony: None declared.Patents: None declared.

Role of Sponsor: The funding organizations played no role in thedesign of study, choice of enrolled patients, review and interpretation ofdata, or preparation or approval of manuscript.

References

1. Gahr P, DeVries AS, Wallace G, Miller C, Kenyon C,Sweet K, et al. An outbreak of measles in an undervac-cinated community. Pediatrics 2014;134:e220 – e8.

2. WHO. WHO-recommended standards for surveillanceof selected vaccine-preventable diseases. Geneva,Switzerland: World Health Organization, 2003.

3. WHO. Rubella. http://www.who.int/mediacentre/factsheets/fs367/en/ (Accessed September 2013).

4. McKenzie KG, Lafleur LK, Lutz BR, Yager P. Rapid proteindepletion from complex samples using a bead-basedmicrofluidic device for the point of care. Lab Chip 2009;9:3543– 8.

5. Champsaur H, Fattal-German M, Arranhado R. Sensitiv-ity and specificity of viral immunoglobulin M determi-nation by indirect enzyme-linked immunosorbent as-say. J Clin Microbiol 1988;26:328 –32.

6. Meurman OH, Ziola BR. IgM-class rheumatoid factor in-terference in the solid-phase radioimmunoassay ofrubella-specific IgM antibodies. J Clin Pathol 1978;31:483–7.

7. Alere(TM). ImmunoComb® rubella IgM. http://www.alere.com/ww/en/product-details/immunocomb-rubella-igm.html (Accessed July 2014).

8. Kiyaga C, Sendagire H, Joseph E, McConnell I, Grosz J,

428 Clinical Chemistry 61:2 (2015)

Narayan V, et al. Uganda’s new national laboratory sam-ple transport system: a successful model for improvingaccess to diagnostic services for early infant HIV diagno-sis and other programs. PLoS One 2013;8:e78609.

9. Reef SE, Strebel P, Dabbagh A, Gacic-Dobo M, Cochi S.Progress toward control of rubella and prevention ofcongenital rubella syndrome: worldwide, 2009. J InfectDis 2011;204:S24 –S7.

10. Ng AHC, Uddayasankar U, Wheeler AR. Immunoassaysin microfluidic systems. Anal Bioanal Chem 2010;397:991–1007.

11. Shriver-Lake L, Golden J, Bracaglia L, Ligler F. Simulta-neous assay for ten bacteria and toxins in spiked clinicalsamples using a microflow cytometer. Anal BioanalChem 2013;405:5611– 4.

12. Phurimsak C, Yildirim E, Tarn MD, Trietsch SJ, Hanke-meier T, Pamme N, Vulto P. Phaseguide assisted liquidlamination for magnetic particle-based assays. LabChip 2014;14:2334 – 43.

13. Martinez AW, Phillips ST, Butte MJ, Whitesides GM.Patterned paper as a platform for inexpensive, low-volume, portable bioassays. Angew Chem Int Ed 2007;46:1318 –20.

14. Pollock NR, Rolland JP, Kumar S, Beattie PD, Jain S,Noubary F, et al. A paper-based multiplexed transami-nase test for low-cost, point-of-care liver function test-ing. Sci Transl Med 2012;4:152ra29.

15. Fu E, Liang T, Spicar-Mihalic P, Houghtaling J, Ram-achandran S, Yager P. Two-dimensional paper networkformat that enables simple multistep assays for use inlow-resource settings in the context of malaria antigendetection. Anal Chem 2012;84:4574 –9.

16. Tekin HC, Gijs MAM. Ultrasensitive protein detection:a case for microfluidic magnetic bead-based assays. LabChip 2013;13:4711–39.

17. Chin CD, Cheung YK, Laksanasopin T, Modena MM,Chin SY, Sridhara AA, et al. Mobile device for diseasediagnosis and data tracking in resource-limited set-tings. Clin Chem 2013;59:629 – 40.

18. Chin CD, Laksanasopin T, Cheung YK, Steinmiller D,Linder V, Parsa H, et al. Microfluidics-based diagnosticsof infectious diseases in the developing world. Nat Med2011;17:1015–9.

19. Jokerst JV, McDevitt JT. Programmable nano-bio-chips: multifunctional clinical tools for use at the point-of-care. Nanomedicine 2009;5:143–55.

20. Lafleur L, Stevens D, McKenzie K, Ramachandran S,

Spicar-Mihalic P, Singhal M, et al. Progress toward mul-tiplexed sample-to-result detection in low resource set-tings using microfluidic immunoassay cards. Lab Chip2012;12:1119 –27.

21. Fridley GE, Holstein CA, Oza SB, Yager P. The evolutionof nitrocellulose as a material for bioassays. MRS Bull2013;38:326 –30.

22. Apilux A, Ukita Y, Chikae M, Chailapakul O, Takamura Y.Development of automated paper-based devices for se-quential multistep sandwich enzyme-linked immu-nosorbent assays using inkjet printing. Lab Chip 2013;13:126 –35.

23. Lutz B, Liang T, Fu E, Ramachandran S, Kauffman P,Yager P. Dissolvable fluidic time delays for program-ming multi-step assays in instrument-free paper diag-nostics. Lab Chip 2013;13:2840 –7.

24. Whitesides GM. Viewpoint on ”Dissolvable fluidictime delays for programming multi-step assays ininstrument-free paper diagnostics.” Lab Chip 2013;13:4004 –5.

25. Choi K, Ng AHC, Fobel R, Wheeler AR. Digital microflu-idics. Annu Rev Anal Chem 2012;5:413– 40.

26. Sista RS, Eckhardt AE, Wang T, Graham C, Rouse JL,Norton SM, et al. Digital microfluidic platform for mul-tiplexing enzyme assays: implications for lysosomalstorage disease screening in newborns. Clin Chem2011;57:1444 –51.

27. Srinivasan V, Pamula VK, Fair RB. An integrated digitalmicrofluidic lab-on-a-chip for clinical diagnostics on hu-man physiological fluids. Lab Chip 2004;4:310 –5.

28. Cho SK, Moon H, Kim C-J. Creating, transporting, cut-ting, and merging liquid droplets by electrowetting-based actuation for digital microfluidic circuits. J Micro-electromech Syst 2003;12:70 – 80.

29. Fobel R, Kirby AE, Ng AHC, Farnood RR, Wheeler AR.Paper microfluidics goes digital. Adv Mater 2014;26:2838 – 43.

30. Fobel R, Fobel C, Wheeler AR. DropBot: an open-sourcedigital microfluidic control system with precise controlof electrostatic driving force and instantaneous dropvelocity measurement. Appl Phys Lett 2013;102:193513.

31. Peng C, Zhang Z, Kim CJ, Ju YS. EWOD (electrowettingon dielectric) digital microfluidics powered by fingeractuation. Lab Chip 2014;14:1117–22.

32. Choi K, Ng AHC, Fobel R, Chang-Yen DA, Yarnell LE,Pearson EL, et al. Automated digital microfluidic plat-

form for magnetic-particle-based immunoassays withoptimization by design of experiments. Anal Chem2013;85:9638 – 46.

33. Ng AHC, Choi K, Luoma RP, Robinson JM, Wheeler AR.Digital microfluidic magnetic separation for particle-based Immunoassays. Anal Chem 2012;84:8805–12.

34. Vergauwe N, Vermeir S, Wacker JB, Ceyssens F, Corna-glia M, Puers R, et al. A highly efficient extraction pro-tocol for magnetic particles on a digital microfluidicchip. Sens Actuator B-Chem 2014;196:282–91.

35. Emani S, Sista R, Loyola H, Trenor CCI, Pamula VK,Emani SM. Novel microfluidic platform for automatedlab-on-chip testing of hypercoagulability panel. BloodCoagul Fibrinolysis 2012;23:760 – 8.

36. Banoo S, Bell D, Bossuyt P, Herring A, Mabey D, Poole F,et al. Evaluation of diagnostic tests for infectiousdiseases: general principles. Nat Rev Microbiol 2008;8:S16 –S28.

37. WHO. Operational characteristics of commerciallyavailable assays to determine antibodies to HIV-1 andor HIV-2 in human sera. Report 11. Geneva, 1999.

38. Au SH, Kumar P, Wheeler AR. A new angle on pluronicadditives: advancing droplets and understanding indigital microfluidics. Langmuir 2011;27:8586 –94.

39. Fu E, Yager P, Floriano PN, Christodoulides N, McDevittJT. Perspective on diagnostics for global health. IEEEPulse 2011;2:40 –50.

40. Dryden MD, Rackus DD, Shamsi MH, Wheeler AR. Inte-grated digital microfluidic platform for voltammetricanalysis. Anal Chem 2013;85:8809 –16.

41. Shamsi MH, Choi K, Ng AHC, Wheeler AR. A digital mi-crofluidic electrochemical immunoassay. Lab Chip2014;14:547–54.

42. Wilson D. Rheumatoid factors in patients with rheuma-toid arthritis. Can Fam Phys 2006;52:180 –1.

43. Tipples GA. Rubella diagnostic issues in Canada. J In-fect Dis 2011;204:S659 –S63.

44. Dimech W, Panagiotopoulos L, Francis B, Laven N, Mar-ler J, Dickeson D, et al. Evaluation of eight anti-rubellavirus immunoglobulin G immunoassays that report re-sults in international units per milliliter. J Clin Micro-biol 2008;46:1955– 60.

45. Pham VH, Nguyen TV, Nguyen TTT, Dang LD, Hoang NH,Nguyen TV, Abe K. Rubella epidemic in Vietnam: char-acteristic of rubella virus genes from pregnant womenand their fetuses/newborns with congenital rubellasyndrome. J Clin Virol 2013;57:152– 6.

Microfluidic Rubella Assay

Clinical Chemistry 61:2 (2015) 429